molecules

Article Indole-Containing Phytoalexin-Based Bioisosteres as Antifungals: In Vitro and In Silico Evaluation against Fusarium oxysporum

Andrea Angarita-Rodríguez, Diego Quiroga and Ericsson Coy-Barrera * Bioorganic Chemistry Laboratory, Facultad de Ciencias Básicas y Aplicadas, Universidad Militar Nueva Granada, Campus Nueva Granada, Cajicá 250247, Colombia; [email protected] (A.A.-R.); [email protected] (D.Q.) * Correspondence: [email protected]; Tel.: +57-1-6500000 (ext. 3270)

Academic Editors: Maria Daglia, Simone Carradori and Annabella Vitalone



 Received: 18 November 2019; Accepted: 20 December 2019; Published: 21 December 2019

Abstract: There is a continuous search for more reliable and effective alternatives to control phytopathogens through different strategies. In this context, indole-containing phytoalexins are stimuli-induced compounds implicated in plant defense against plant pathogens. However, phytoalexins’ efficacy have been limited by fungal detoxifying mechanisms, thus, the research on bioisosteres-based analogs can be a friendly alternative regarding the control of Fusarium phytopathogens, but there are currently few studies on it. Thus, as part of our research on antifungal agents, a set of 21 synthetic indole-containing phytoalexin analogs were evaluated as inhibitors against the phyopathogen Fusarium oxysporum. Results indicated that analogs of the N,N-dialkylthiourea, N,S-dialkyldithiocarbamate and substituted-1,3-thiazolidin-5-one groups exhibited the best docking scores and interaction profiles within the active site of Fusarium spp. enzymes. Vina scores exhibited correlation with experimental mycelial growth inhibition using supervised statistics, and this antifungal dataset correlated with molecular interaction fields after CoMFA. Compound 24 (tert-butyl (((3-oxo-1,3-diphenylpropyl)thio)carbonothioyl)-l-tryptophanate), a very active analog against F. oxysporum, exhibited the best interaction with lanosterol 14α-demethylase according to molecular docking, molecular dynamics and molecular mechanic/poisson-boltzmann surface area (MM/PBSA) binding energy performance. After data analyses, information on mycelial growth inhibitors, structural requirements and putative enzyme targets may be used in further antifungal development based on phytoalexin analogs for controlling phytopathogens.

Keywords: phytopathogens; Fusarium spp.; antifungals; indole-containing analogs

1. Introduction Fungi from the genus Fusarium are very important because they have wide cosmopolitan-type geographic distribution and high relevance for agriculture and economy [1]. Their diversity is quite common in soils and many of them have the capacity to cause infectious diseases in different types of crop plants. Additionally, some species can cause opportunistic infections in animals due to the production of toxins that can affect them [1]. Genus Fusarium gathered phytopathogenic-behaved species causing necrosis-associated diseases in several plants around the world. This fungus can survive in the soil in mycelial or conidial forms during host absence. When a host is present, the mycelium penetrates its roots, enters the vascular system (xylem) to move and multiply itself, causing wilt symptoms. Soil-borne fungi Fusarium spp. can infects the host by using different sets of genes for early host-plant signaling, as well as the binding and enzymatic decomposition of antifungal compounds produced by plant for defense, in order to inactivate and kill host cells by

Molecules 2020, 25, 45; doi:10.3390/molecules25010045 www.mdpi.com/journal/molecules Molecules 2020, 25, 45 2 of 19 fungal toxins [2]. In this context, some control strategies based on chemical, physical and cultural methods have arisen to mitigate the negative effects of this phytopathogen to host plants. Chemical treatments include formaldehyde applications to avoid disseminations as well as dazomet, sodium methan, methyl isothiocyanate and systemic fungicides as benomyl, thiabendazon, carbendazim and methylthiophanate [3]. Fusarium-caused diseases treatment by systemic fungicides, although it has good effectivity, is also problematic, since these kind of treatments can become mutagenics for several plants and these agents can also generate high degrees of resistance [3,4]. Plants involve a systemic defense mechanism involving some metabolites such as phytoalexins. These metabolites are synthesized in adjacent areas of healthy cells to those damaged cells and they are accumulated both in necrotic and susceptible resistant tissues [5]. In other words, they are strictly produced in a site around the place where infection needs to be controlled. Thus, the resistance occurs when one or more phytoalexins reach a high enough concentration to inhibit the pathogen development [5]. However, some phytopathogens are recently described to exhibit strategies to invade the plant tissues and obtain necessary nutrients for its growing and reproduction [1,2,4]. Despite that plants are capable to synthesize specific antifungal metabolites, the fungi are capable to produce particular enzymes in order to detoxify quickly and inhibit the phytoalexin activity. This detoxifying capacity is described by Pedras et al. [6] as an “arms race” that advantaged the pathogen, causing adverse effects in the crop. An example of the previous fact is the case of brassinin (an important phytoalexin for crucifers due its dual potential as an antimicrobial defense and a biosynthetic precursor of phytoalexin variety), since if this metabolite is inhibited by a fungal enzyme, the plant become possibly more susceptible to be attacked by the phytopathogen [6]. Thus, the cruciferous invader L. maculans (a devastating pathogen) can generate itself a detoxification defense against phytoalexins. The mechanism is based on brassinin oxidase production, which catalyzes the brassinin oxidation through the dithiocarbamate transformation into an aldehyde. This mechanism depends on direct brassinin recognition by this enzyme. Therefore, exploration of brassinin analogs acting as bioisosteres can be adopted as an important approach for the development of novel antifungals [7]. In this sense, bioisosterism is generally a chemical strategy for the rational design of bioactives from structural modifications of leads, whose action mechanism and chemical structure is therefore known. From the above fact, it is possible to develop novel compounds with chemical and/or physical similarities but producing identical or even better biological properties [7]. Therefore, as part of our research on antifungals discovery and development, the present work was initially focused on the in silico exploration, by molecular docking, on the interaction of a set of indole-containing phytoalexin analogs within the active site of 25 enzymes reported to Fusarium spp. Subsequently, the integration of the experimental antifungal activity (i.e., Fusarium oxysporum mycelial growth inhibition) and docking scores datasets (by supervised multivariate statistics) as well as the three-dimensional (3D) molecular descriptors and antifungal activity datasets (on the basis of 3D quantitative structure-activity relationship (3D-QSAR) by comparative molecular field analysis (CoMFA)) led to recognize some important hits, enzyme targets and structural requirements to be stated for the further development of antifungals for controlling phytopathogens using phytoalexin-like analogs. Finally, some insights into the binding mode of a 3-oxo-1,3-diphenylpropyl N-alkyldithiocarbamate-type analog and the homology model of the transmembrane F. cerealis lanosterol 14-demethylase were also determined in order to rationalize the antifungal activity of this type of synthetic phytoalexin analogs.

2. Results and Discussion

2.1. Test Indole-Containing Phytoalexin Analogs and Docking Protocol Validation A set of 25 synthetic indole-containing compounds (1–25) (Figure1) were obtained by our previously reported reaction route from l-tryptophan [8] (Figure S1). Their IUPAC names are displayed in the complementary material (Table S1). Such compounds were then chosen as test compounds owing to their antifungal activity by inhibiting mycelial growth of F. oxysporum [8,9]. Thus, in order to explore Molecules 2020, 25, 45 3 of 19 Molecules 2019, 24, x FOR PEER REVIEW 3 of 19 andThus, outline in order a further to explore information and outline related a to further the potential information and putative related targets to the of potential these compounds and putative at in silicotargets level, of these docking compounds simulations at in were silico therefore level, do performedcking simulations with 25 fungalwere therefore enzymes performed (E1–E25) (Table with A1 25 (Appendixfungal enzymesA)), which (E1–E25) were (Table selected A1), due which to their were significant selected associations due to their on significant pathogenic associations and metabolic on performancepathogenic and of Fusarium metabolicspp. performance of Fusarium spp.

FigureFigure 1. 1. StructuralStructural classification classification of the of test the indo testle-containing indole-containing phytoalexin phytoalexin analogs: analogs:2-alkyl 2-alkylaminoesters aminoesters (1–4), N (,1N–4-dialkylthioureas), N,N-dialkylthioureas (5–8), 2-cyanoethyl (5–8), 2-cyanoethyl N-alkyldithiocarbamatesN-alkyldithiocarbamates (9–12), (3-methoxy-3-oxopropyl9–12), 3-methoxy-3-oxopropyl N-alkyldithiocarbamateN-alkyldithiocarbamate (13 (–1316–),16 ),2-methyl-4-oxopentan-2-yl NN-alkyldithiocarbamate-alkyldithiocarbamate ((1717––2020),), 2-oxo-1,3-diphenylpropyl 2-oxo-1,3-diphenylpropyl N-alkyldithiocarbamates-alkyldithiocarbamates ( (21–24), 4-[(14-[(1HH-indol-3-yl)-indol-3-yl) methylene]-2-sulfanilidene-1,3-thiazolidin-5-one methylene]-2-sulfanilidene-1,3-thiazolidin-5-one (25 (),25 brassinin), brassinin (26 ).(26).

AmongAmong the several in in silico silico approaches approaches for for the the stud studyy of of active active chemical chemical agents agents from from public public or orin-house in-house libraries libraries [10], [10 ],molecular molecular docking docking was was then then selected selected due due to to the the capacity capacity to simulatesimulate thethe bindingbinding modemode ofof low-molecularlow-molecular weightweight compoundscompounds within the activeactive sitesite ofof enzymesenzymes andand bindingbinding pocketpocket ofof receptors.receptors. This is a veryvery importantimportant procedure for thethe predictionprediction ofof putativeputative targets,targets, asas thethe first-linefirst-line stepstep inin thethe searchsearch forfor plausibleplausible mechanismmechanism ofof actionaction ofof activeactive compoundscompounds [[11].11]. However, somesome problemsproblems/issues/issues can can arise arise and and they they should should be takenbe taken into accountinto account for caution for caution and concern and concern during structure-basedduring structure-based screening screening [12], as well[12], asas thewell selection as the selection of suitable of suitable protocol protocol for sampling for sampling and scoring and procedurescoring procedure to discriminate to discriminate potential binderspotential from bind non-bindersers from non-binders using accurate usin parametersg accurate [parameters13]. In this regard,[13]. In this the molecularregard, the docking molecular of docking the co-crystalized of the co-crystalized ligands (CCL) ligands for each(CCL) enzyme for each let enzyme us evaluate let us theevaluate good performancethe good performance of the docking of the protocol docking used, prot sinceocol thisused, re-docking since this calculations re-docking resulted calculations into conformationalresulted into conformational root-mean-square root-mean-square deviation (RMSD) deviation values <(RMSD)1.0, indicating values < that 1.0, our indicating parameters that were our suitableparameters for were docking. suitable Additionally, for docking. this Additionally, performance this was performance also evaluated was throughalso evaluated a benchmarking through a strategybenchmarking using a strategy set of decoys using (n a= set1125) of decoys and active (n = compounds 1125) and (activen = 55) compounds against five ( enzymesn = 55) against within five test α setenzymes (i.e., isocitrate within lyase, test aspartateset (i.e., kinase, isocitrate sereine lyase, esterase, aspartate lanosterol kinase, 14 -demethylase sereine esterase, and trichothecene lanosterol acetyltransferase)14α-demethylase havingand trichothecene known inhibitors, acetyltransferas retrievede) from having literature known and inhibitors, databases. retrieved The sensitivity from andliterature specificity and ofdatabases. the docking The protocol sensitivity was assessedand spec byificity calculation of the docking of the area protocol under curve was (AUC)assessed from by receivercalculation operating of the area characteristic under curve (ROC) (AUC) curves from [14 rece]. Thisiver assessment operating characteristic of our docking (ROC) protocol curves reached [14]. trueThis positiveassessment identification of our docking over 90% protocol of the activereached compounds true positive involving identification a recognition over within90% of 10%the active of the testcompounds compounds involving (Figure 2aa). recognition Therefore, thewithin validation 10% of of the the test docking compounds protocol (Figure was considered 2a). Therefore, successful, the sincevalidation the AUC of andthe thedocking Boltzmann-enhanced protocol was discriminationconsidered successful, of the receiver since operating the AUC characteristic and the (BEDROC)Boltzmann-enhanced fell into 0.931–0.961 discrimination and 0.819–0.886 of the receiv ranges,er respectivelyoperating characteristic (Figure2b). Previously, (BEDROC) such fell a kindinto of0.931–0.961 validation and let to0.819–0.886 the discovery ranges, of new respectively chemotypes (Figur fore inhibition 2b). Previously, of homoserine such a kind dehydrogenase of validation [15 let], asto anthe indication discovery of theof new importance chemotyp of validatinges for inhibition the docking of homoseri protocols.ne dehydrogenase [15], as an indication of the importance of validating the docking protocols. Once the parameters were adequately validated, each Merck molecular force field (MMFF)-optimized structure 1–26 was docked into the active site of each enzyme using

Molecules 2019, 24, x FOR PEER REVIEW 4 of 19

AutoDock/Vina. The calculated affinities (as Vina scores, expressed in kcal/mol) of the resulting enzyme-ligand complexes, for each enzyme and compound, were compiled as a data matrix summarized in Table S2. This dataset revealed some trends around the affinity values as a result of the simulated molecular interaction between analogs and enzyme targets, through the use of descriptive and multivariate analysis. For this purpose, test compounds 1–25 were subdivided into seven classes according to their moieties: alkyl 2-aminoesters (AAE) 1–4, N,N-dialkylthioureas (NNDATU) 5–8, 2-cyanoethyl N-alkyldithiocarbamates (CE-DTC) 9–12, 3-methoxy-3-oxopropyl N-alkyldithiocarbamate (MOPr-DTC) 13–16, 2-methyl-4-oxopentan-2-yl N-alkyldithiocarbamate (MOPe-DTC) 17–20, 2-oxo-1,3-diphenylpropyl N-alkyldithiocarbamates (ODP-DTC) 21–24 and 4-[(1MoleculesH-indol-3-yl)-methylene]-2-sulfanyl2020, 25, 45 idene-1,3-thiazolidin-5-one (IST) 25, along with brassinin4 of 19 (26) as the reference phytoalexin [6].

FigureFigure 2. 2.BenchmarkingBenchmarking of of the the docking docking protocol protocol using aa setset ofof active active compounds compounds decoys decoys against against test test enzymesenzymes (red (red line line = isocitrate= isocitrate lyase; lyase; green green line line = aspart= aspartateate kinase; kinase; black black line line = seri= serinene esterase; esterase; purple purple line = lanosterolline = lanosterol 14α-demethylase; 14α-demethylase; dark orange dark line orange = tr lineichothecene= trichothecene acetyltransferase). acetyltransferase). (a) Receiver (a) Receiver operating characteristicoperating characteristic (ROC). (b) Enrichment (ROC). (b) curve. Enrichment curve.

2.2. VinaOnce Scores-Related the parameters Trends were adequately validated, each Merck molecular force field (MMFF)-optimized structure 1–26 was docked into the active site of each enzyme using AutoDock/Vina. The affinity values for each ligand-enzyme complexes were compiled into a full data matrix. A The calculated affinities (as Vina scores, expressed in kcal/mol) of the resulting enzyme-ligand reduced dataset expressed as mean Vina scores ± standard deviation (SD)—after 10 independent complexes, for each enzyme and compound, were compiled as a data matrix summarized calculations—is exposed in Table S2. The SD for Vina scores exhibited good performance between in Table S2. This dataset revealed some trends around the affinity values as a result of replicates, having coefficient of variation into the 5–15% range, although with some outliers. The the simulated molecular interaction between analogs and enzyme targets, through the use of root-mean-square deviation (RMSD) of atomic positions for the best-ranked poses between descriptive and multivariate analysis. For this purpose, test compounds 1–25 were subdivided replicates exhibited excellent convergence (<1.0 Å). The lowest scores per enzyme and ligand were into seven classes according to their moieties: alkyl 2-aminoesters (AAE) 1–4, N,N-dialkylthioureas highlighted in red. From these data, ODP-DTC, NNDATU and IST-type ligands were found to (NNDATU) 5–8, 2-cyanoethyl N-alkyldithiocarbamates (CE-DTC) 9–12, 3-methoxy-3-oxopropyl generate more stable ligand-enzyme complexes with 13, eight and four enzymes, respectively. These N-alkyldithiocarbamate (MOPr-DTC) 13–16, 2-methyl-4-oxopentan-2-yl N-alkyldithiocarbamate resulting Vina scores were analyzed by Pearson’s correlation (Table 1). Thus, the respective (MOPe-DTC) 17–20, 2-oxo-1,3-diphenylpropyl N-alkyldithiocarbamates (ODP-DTC) 21–24 and coefficients were independently calculated between test compounds as well as among docked 4-[(1H-indol-3-yl)-methylene]-2-sulfanylidene-1,3-thiazolidin-5-one (IST) 25, along with brassinin enzymes. Firstly, the Pearson’s coefficients (Pc) for some pair of compound classes, according to (26) as the reference phytoalexin [6]. their complete dataset of Vina scores for all enzymes, exhibited values above 0.9. This correlation indicated2.2. Vina a Scores-Relatedsimilar interaction Trends profile with the enzyme group. Therefore, AAE exhibited similar performance with MOPe-DTC and IST, whereas MOPr-DTC with AAE,The aCE-DTC,ffinity values MOPe-DTC for each and ligand-enzyme IST. NNDATU complexes and ODP-DTC were compiled exhibited into coefficient a full data values matrix. A reduced dataset expressed as mean Vina scores standard deviation (SD)—after 10 independent below 0.75 and 0.65, respectively, suggesting different± trends in the molecular docking results in comparisoncalculations—is to the exposedother compound in Table S2.classes. The Regard SD foring Vina the scores Pc across exhibited enzymes good (Table performance S3), 13 paired between correlationsreplicates, were having detected coeffi cientwith values of variation above into 0.8, thewhich 5–15% implies range, similar although interaction with someprofile outliers. according The to root-mean-squarethe structural variations deviation of test (RMSD) compound of atomic set and positions therefore for thedepending best-ranked poss posesibly by between the size, replicates shape and/orexhibited chemical excellent behavior convergence of the active (<1.0 site Å). Theof each lowest enzyme scores and per even enzyme enzyme and ligand family. were For highlightedinstance, thein pairs red. E2–E3 From theseand E18–E19 data, ODP-DTC, were found NNDATU to be correlated and IST-type (Pc = ligands 0.879 and were 0.812, found respectively) to generate and more theystable belong ligand-enzyme to the same enzyme complexes families, with 13,i.e.,eight glycosyl and hydrolase four enzymes, and nitroalkane respectively. oxidase These families, resulting Vina scores were analyzed by Pearson’s correlation (Table1). Thus, the respective coe fficients were

independently calculated between test compounds as well as among docked enzymes. Firstly, the Pearson’s coefficients (Pc) for some pair of compound classes, according to their complete dataset of Vina scores for all enzymes, exhibited values above 0.9. This correlation indicated a similar interaction profile with the enzyme group. Molecules 2020, 25, 45 5 of 19

Table 1. Values of Pearson correlation for subsets of test compounds.

CC a C1 C2 C3 C4 C5 C6 C7 C1 1.00 C2 0.699 1.00 C3 0.872 0.832 1.00 C4 0.926 0.731 0.901 1.00 C5 0.921 0.747 0.886 0.985 1.00 C6 0.647 0.626 0.627 0.678 0.724 1.00 C7 0.928 0.704 0.886 0.912 0.888 0.636 1.00 C8 0.728 0.444 0.670 0.693 0.671 0.446 0.694 a Compound classes: C1: alkyl 2-aminoesters (AAE); C2: N,N-dialkylthioureas (NNDATU); C3: 2-cyanoethyl N-alkyldithiocarbamates (CE-DTC); C4: 3-methoxy-3-oxopropyl N-alkyldithiocarbamate (MOPr-DTC), C5: 2-methyl-4-oxopentan-2-yl N-alkyldithiocarbamate (MOPe-DTC); C6: 2-oxo-1,3-diphenylpropyl N-alkyldithiocarbamates (ODP-DTC), C7: 4-[(1H-indol-3-yl)-methylene]-2-sulfanylidene-1,3-thiazolidin-5-one (IST); C8: brassinin.

Therefore, AAE exhibited similar performance with MOPe-DTC and IST, whereas MOPr-DTC with AAE, CE-DTC, MOPe-DTC and IST. NNDATU and ODP-DTC exhibited coefficient values below 0.75 and 0.65, respectively, suggesting different trends in the molecular docking results in comparison to the other compound classes. Regarding the Pc across enzymes (Table S3), 13 paired correlations were detected with values above 0.8, which implies similar interaction profile according to the structural variations of test compound set and therefore depending possibly by the size, shape and/or chemical behavior of the active site of each enzyme and even enzyme family. For instance, the pairs E2–E3 and E18–E19 were found to be correlated (Pc = 0.879 and 0.812, respectively) and they belong to the same enzyme families, i.e., glycosyl hydrolase and nitroalkane oxidase families, respectively. This outcome suggested good performance of the docking calculations in accordance with the good convergence found in the output data. Additionally, E5 showed good correlation with E2 and E3 (Pc > 0.75), but E5 is part of a different enzyme family, i.e., transferase that catalyzes the transfer of an acetyl group to the hydroxyl at C3 of several trichothecene-type mycotoxins [16]. In this sense, it is reasonable to consider that such a tendency is owing to the presence of similar docking interactions between residues from active site of different enzymes and the compound set. In Table S3 are highlighted those correlated enzyme pairs from resulting Vina scores after docking that support these observations. In order to observe the particular variation/distribution of Vina scores between enzymes and test ligands, the respective box plots were also constructed using the whole dataset (Figure3). Thus, 12 enzymes exhibited good reproducibility and low dispersion in the docking calculations for six types of ligands, whereas AAE and CE-DTC showed more data scattering. The best behavior was exhibited by ODP-DTC-type phytoalexin, excepting for E24, because two outliers were produced for 21 and 22. A similar dispersion was also observed for other compound type on interacting with specific enzymes, indicating precise structural requirements to interact depending on the alkyl group at ester moiety. Concerning the best-docked complexes, MOPe-DTC and CE-DTC-type ligands were associated to a best interaction with E18, whereas MOPr-DTC-type analogs might inhibit putatively the enzymes E18 and E24. ODP-DTC-type compounds showed the best Vina scores with E2, E17, E18 and E25. Regarding brassinin, its behavior was found to be comparable to that of IST. These facts led to infer that significant differences in the structural interactions between enzymes and test ligands would influence the capacity to form a stable complex. Molecules 2020, 25, 45 6 of 19 Molecules 2019, 24, x FOR PEER REVIEW 6 of 19

FigureFigure 3. 3. BoxplotsBoxplots for foraffinity affinity values values (as Vina (as Vinascores scores in kcal/mol) in kcal after/mol) molecular after molecular docking dockingfor test phytoalexinfor test phytoalexin analogs. (a analogs.) 2-alkyl aminoesters (a) 2-alkyl (AAE), aminoesters (b) N,N (AAE),-dialkylthioureas (b) N,N -dialkylthioureas(NNDATU), (c) 2-cyanoethyl(NNDATU), (c) 2-cyanoethylN-alkyldithiocarbamatesN-alkyldithiocarbamates (CE-DTC), (CE-DTC), (d) (d)3-methoxy-3-oxopropyl NN-alkyldithiocarbamate-alkyldithiocarbamate (MOPr-DTC), (MOPr-DTC), ( (e)) 2-methyl-4-oxopentan-2-yl2-methyl-4-oxopentan-2-yl N-alkyldithiocarbamate (MOPe-DTC),(MOPe-DTC), (f) 2-oxo-1,3-diphenylpropyl(f) 2-oxo-1,3-diphenylpropylN-alkyldithiocarbamates N-alkyldithiocarbamates (ODP-DTC), (g(ODP-DTC),) 4-[(1H-indol-3-yl) (g) 4-[(1methylene]-2-sulfanilidene-1,3-thiazolidin-5-oneH-indol-3-yl) methylene]-2-sulfanilidene-1,3-thiazolidin-5-one (IST), (h) brassinin. (IST), (h) brassinin.

2.3.2.3. Unsupervised Unsupervised and and Supervi Supervisedsed Multivariate Multivariate Statistics Statistics TheThe plausible plausible relationships, relationships, from from a a holistic holistic point-of-view, point-of-view, were studied through multivariate analysisanalysis using using the the Vina Vina scores dataset for each enzyme (as variables) and test compounds/natural compounds/natural ligandsligands (as(as observations).observations). Hence, Hence, an unsupervisedan unsupervised exploration exploration through through principal principal component component analysis analysis(PCA) was (PCA) firstly was started firstly and started the resulting and the PC1 resulting versus PC1 PC2 versus was obtained PC2 was (Figure obtained S2). As(Figure expected, S2). theAs expected,docking results the docking of natural results ligands of natural (NL) showed ligands a (NL) different showed behavior a different to that behavior of those Vina to that scores of those of all Vinatest analogs scores of (Figure all test S2a); analogs thus, (Figure they were S2a); excluded thus, they in orderwere excluded to start an in additional order to start PCA. an Figure additional S2b,c PCA.show Figures the resulting S2b and score S2c plot show (uncolored the resulting and colored score byplot compound (uncolored type, and respectively), colored by compound which exhibits type, a respectively),clustering depending which exhibits on the di aff clusteringerent classes depending of compounds on the within different dataset. classes This of examination compounds was within then dataset.refined usingThis examination a supervised was examination then refined by using means a ofsupervised Orthogonal examination Partial Least by Squares-Discriminantmeans of Orthogonal PartialAnalysis Least (OPLS-DA) Squares-Discriminant using the compound Analysis type(OPLS-DA) as categorical using the variable, compound in order type toas observecategorical the variable,particular in connection order to observe of the dockingthe particular scores connection related to theof the structures docking of scores test analogs. related to In the summary, structures the ofOPLS-DA-derived test analogs. In summary, score plot th definede OPLS-DA-derived more clearly that score NNDATU, plot defined IST andmore ODP-DTC clearly that were NNDATU, clustered ISTas independent and ODP-DTC groups were in comparisonclustered as to independent the other compound groups in type comparison (Figure4a) to as evidencedthe other compound in Pearson typecorrelations (Figure and4a) as box evidenced plots distributions. in Pearson correlations and box plots distributions.

Molecules 2020, 25, 45 7 of 19 Molecules 2019, 24, x FOR PEER REVIEW 7 of 19

Figure 4. 4. SupervisedSupervised multivariate multivariate analysis analysis from Vina from scores Vina for scoresindole-containing for indole-containing phytoalexin phytoalexinanalogs. (a) Orthogonal analogs. partial (a) least Orthogonal squares – discrimin partial leastant analysis squares (OPLS-DA)-derived – discriminant score analysis plot. (OPLS-DA)-derivedColoration according scorecompound plot. type: Coloration alkyl 2-aminoesters according (AAE), compound N,N-dialkylthioureas type: alkyl 2-aminoesters (NNDATU), (AAE),2-cyanoethylN,N -dialkylthioureasN-alkyldithiocarbamates (NNDATU), (CE-DTC), 2-cyanoethyl 3-methoxy-3-oxopropylN-alkyldithiocarbamates N-alkyldithiocarbamate (CE-DTC), 3-methoxy-3-oxopropyl(MOPr-DTC), 2-methyl-4-oxopentan-2-ylN-alkyldithiocarbamate N (MOPr-DTC),-alkyldithiocarbamate 2-methyl-4-oxopentan-2-yl (MOPe-DTC), N2-oxo-1,3-diphenylpropyl-alkyldithiocarbamate (MOPe-DTC), 2-oxo-1,3-diphenylpropylN-alkyldithiocarbamatesN-alkyldithiocarbamates (ODP-DTC), (ODP-DTC), 4-[(1H-indol-3-yl)-methylene]-2-sulfanyliden-indol-3-yl)-methylene]-2-sulfanylidene-1,3-thiazolidin-5-onee-1,3-thiazolidin-5-one (IST); (IST); ( b) OPLS-DA-derivedOPLS-DA-derived loadings plot; (c) Variable importanceimportance inin projectionprojection (VIP)(VIP) scoresscores plot.plot. X-scores (t[1] or t[2]) in (a) and XY-combined loadingsloadings (pq[1] or pq[2]) in (b)(b) werewere proportionallyproportionally scaledscaled toto thethe explainedexplained variationvariation 2 (R(R2X)) using using a a coefficient coefficient as as multiplicative multiplicative factor factor (*). (*).

The respective loadings plot (Figure4 4b)b) exposedexposed thosethose variablesvariables (i.e.,(i.e., enzymes)enzymes) thatthat influenceinfluence the observations (i.e., test analogs) discriminati discriminationsons across the Vina scores. In In this this regard, regard, NNDATU, IST and ODP-DTCODP-DTC were reasonably influencedinfluenced by the resulting docking scores with the E14, E21, E24 and E25, which are relatedrelated to thethe lyase,lyase, dehydrogenase,dehydrogenase, and oxidoreductase enzyme families, respectively. TheThe influence influence contribution contribution of variablesof variables across across dataset dataset within within the OPLS-DA the OPLS-DA model [3 model+3+0] was[3+3+0] estimated was estimated by the variable by the importance variable importance in projection in (VIP)projection scores. (VIP) E25 showedscores. E25 the mostshowed deviated the most VIP score,deviated indicating VIP score, probable indicating good selectivity probable ongood the contributionselectivity on particularly the contribution due to lowest particularly docking due scores to forlowest some docking specific scores compound for some types specific (such compound as ODP-DTC), types whereas(such as ODP-DTC), E21 exhibited whereas the highest E21 exhibited one, i.e., poorestthe highest selectivity one, i.e., by poorest some compound selectivity datasetby some (such compound as NNDATU). dataset Particularly,(such as NNDATU). NNDATU Particularly, possessed severalNNDATU chemical possessed and biologicalseveral chemical properties and that biological are directly properties related tothat their are molecular directly structure.related to Theirtheir observedmolecular distructure.fferences areTheir due observed to their substitutionsdifferences ar ande due such to specifictheir substitutions conformations and they such can specific adopt aroundconformations the thioamide they can moiety adopt [ 17around]. the thioamide moiety [17]. In order to to estimate estimate such such tendencies tendencies experiment experimentallyally and and compare compare them them with with the the performance performance of ofthe the present present docking docking study, study, the the mycelial mycelial growth growth inhibition inhibition was was evaluated evaluated for for all compound set against F.F. oxysporum oxysporum. Such. Such evaluation evaluation was was built built on our on prior our studyprior tostudy expand to theexpand search the for search promising for antifungalpromising agentsantifungal [8] andagents the [8] resulting and the half-maximal resulting half-maximal inhibitory concentrations inhibitory concentrations (IC50) are exposed (IC50) are in Tableexposed S4. in Test Table compounds S4. Test exhibitedcompound mycelials exhibited growth mycelial inhibition growth at di ffinhibitionerent levels, at butdifferent some levels, NNDATU, but OPD-DTCsome NNDATU, and CE-DTC OPD-DTC exhibited and CE-DTC the lowest exhibited IC50 values the lowest (<0.8 IC mM),50 values thus, (<0.8 they mM), can bethus, considered they can asbe theconsidered best antifungals as the best within antifungals the experimental within the dataset. experimental A clear structure-activitydataset. A clear structure-activity relationship was notrelationship noticeably was evidenced; not noticeably thus, theevidenced; antifungal thus, data the was antifungal then used data for was supervised then used analysis for supervised in order toanalysis integrate in order the docking to integrate and biological the docking activity and datasets. biological Hnece, activity the datasets. supervised Hnece, analysis the forsupervised docking scoresanalysis by for single- dockingY OPLS scores regression, by single- usingY OPLS the experimental regression, ICusing50 values the experimental as continuous IC variables50 values (i.e., as continuous variables (i.e., Y) (Figure 5), clustered similarly the observations into some groups according to the compound type under the corresponding regression (Figure 5a,b).

Molecules 2020, 25, 45 8 of 19

Y) (Figure5), clustered similarly the observations into some groups according to the compound type Moleculesunder the 2019 corresponding, 24, x FOR PEER regressionREVIEW (Figure5a,b). 8 of 19

Figure 5.5. SupervisedSupervised multivariate multivariate analysis analysis using using Single- Single-Y OPLSY OPLS on Vina on scores Vina datascores (as X)data and (as antifungal X) and

antifungalactivity (half-maximal activity (half-maximal inhibitory concentrations inhibitory conc (IC50entrationsin mM) as (IC Y).50 ( ain) Single-Y mM) as OPLS-derived Y). (a) Single-Y score plot.OPLS-derived Coloration score according plot. heatmapColoration of according IC50 dataset heatmap correlation; of IC (b50) Single-Ydataset correlation; OPLS-derived (b) score Single-Y plot. OPLS-derivedColoration by score compound plot. Coloration type; (c) Single-Yby compound OPLS-derived type; (c) Single-Y loadings OPLS-derived plot; (d) Variable loadings importance plot; (d) Variablein projection importance (VIP) scores in projection plot. X and(VIP) orthogonal scores plot. scores X (t[1]and ororthogonal to[1], respectively) scores (t[1] in (ora,b )to[1], and XY-combinedrespectively) andin ( orthogonala) and (b loadings) and XY-combined (pq[1] or poso[1], and respectively) orthogonal in loadings (c) were proportionally(pq[1] or poso[1], scaled 2 respectively)to the explained in ( variationc) were proportionally (R X) using a coescaledfficient to the as multiplicativeexplained variation factor (R (*).2X) using a coefficient as multiplicative factor (*). Once again, the well-defined, separated groups were NNDATU, OPD-DTC and IST. These results can rationalizedOnce again, sincethe well-defined, NNDATU and separated OPD-DTC groups exhibited were similarNNDATU, docking OPD-DTC profile and between IST. These them, whereasresults can IST rationalized was slightly since different; NNDATU however, and these OPD- threeDTC kind exhibited of compounds similar were docking very distinctprofile between from the them,others. whereas This fact IST represented was slightly a clear different; trend between howeve dockingr, these scores three performance kind of compounds and the experimental were very distinctIC50 values from based the others. on compound This fact type,represented justifying a clear conceivable trend between structure-activity docking scores and performance structure-affi nityand therelationships. experimental Furthermore, IC50 values based the di onfferential compound variables type, (i.e.,justif enzymes)ying conceivable were identified structure-activity through and the structure-affinityloadings. The respective relationships. loadings plotFurthermore, (Figure5c) showedthe differential a remarkable variables influence (i.e., of E14,enzymes) E17, E18, were E20 identifiedand E21 for through the separation the loadings. of NNDATU-based The respective clusters, loadings while plot IST and(Figure OPD-DTC 5c) showed were most a remarkable influenced influenceby E24 and of E25,E14, respectively.E17, E18, E20 Therefore, and E21 for these the resultsseparation are inof accordanceNNDATU-based with the clusters, respective while box-plots IST and (FigureOPD-DTC2) using were only most the influenced Vina scores by performance E24 and forE25, each respectively. compound Therefore, type. Twelve these enzymes results revealed are in VIPaccordance scores with1.0 underthe respective IC50 supervision, box-plots thus, (Figure they 2) can using be consideredonly the Vina important scores performance in the givenmodel for each to ≥ compoundintegrate in type. silico Twelve docking enzymes scores and revealed in-vitro VIP antifungal scores ≥ 1.0 activity under (Figure IC50 supervision,5d). In this thus, regard, they E18 can was be consideredthe most influencing important variable, in the given and model E25 persisted to integrate as the in mostsilico deviated docking influencingscores and in-vitro enzyme. antifungal This last activityfact is an (Figure indicative 5d). ofIn athis particular regard, selectivityE18 was the of most some influencing compounds variable, (i.e., OPD-DTC-type) and E25 persisted to interact as the withmost thisdeviated enzyme influencing (E25). enzyme. This last fact is an indicative of a particular selectivity of some compoundsN,S-dialkyl (i.e., dithiocarbamatesOPD-DTC-type) to have interact already with been this described enzyme (E25). as potential fungicides, specifically as mycelialN,S-dialkyl growth dithiocarbamates inhibitors of several have pathogenic already been fungi de [9scribed,18], but asN potential-alkyl dithiocarbamate fungicides, specifically salts are aslargely mycelial commercialized growth inhibitors as fungicides, of several since pathogenic they have multi-site fungi [9,18], action but [19 N].-alkyl In general, dithiocarbamate dithiocarbamates salts arebecome largely toxic commercialized when they are metabolized as fungicides, by the targetsince fungus,they have especially multi-site on producing action the[19]. isothiocyanate In general, dithiocarbamatesradical. It can be captured, become e.g.,toxic by when sulfhydryl they groupsare me intabolized amino acids by the and target enzymes; fungus, thus, theespecially enzymatic on producingactivity is inhibitedthe isothiocyanate together withradical. subsequent It can be biological captured, disruptions,e.g., by sulfhydryl such as groups lipid metabolism, in amino acids cell andmembrane enzymes; permeability thus, the and enzymatic respiration activity and ATP is production,inhibited together among others with [20subsequent]. biological disruptions, such as lipid metabolism, cell membrane permeability and respiration and ATP production, among others [20].

2.4. Comparative Molecular Field Analysis (CoMFA)

Molecules 2020, 25, 45 9 of 19

2.4. Comparative Molecular Field Analysis (CoMFA) Owing to the trend observed in supervised multivariate statistics (i.e., single-Y OPLS), a three-dimensional quantitative structure-activity relationship (3D-QSAR) study (through a comparative molecular field analysis (CoMFA)) was implemented to correlate structural features and antifungal activity, in order to explain and understand consistently the antifungal activity for test compounds according to steric and electrostatic properties [21]. For this, the best-docked pose of compound 24 was used as template scaffold to define tethers and superimpose the test compounds 1–24. Experimental antifungal data (as logarithmic half-maximal inhibitory concentrations (pIC50)) and aligned structures were subdivided into a training set (70%) to generate the CoMFA model and a test set (30%) for subsequent external validation [21]. The molecular interaction fields (MIF) were then calculated using probes (steric and electrostatic ones), and, after several steps of data pretreatment, partial least squares (PLS) regression was carried out (using up to five PLS components) in order to build linear relationships between variations in the MIF values as a function of changes in the experimental pIC50. After that, the best model required three PLS components, suggesting good correlation between MIF values and experimental pIC50 of test compounds. The resulting regression also afforded the following parameters: correlation coefficient r2 = 0.855, a cross-validated leave-one-out (LOO) coefficient q2 = 0.779, a cross-validated leave-many-out (LMO) coefficient q2 = 0.709 and F-test = 178.489, demonstrating excellent determination coefficient for a statistically predictive, robust model [22]. Observations were randomly organized to perform a Y-scrambling protocol [23] using 30 scramblings and 10 runs. The 2 2 resulting model declined severely (Rscr and Qscr < 0.4), and no correlation was then evidenced, indicating the model was not achieved as a result of a chance correlation. Thus, according to this comparative analysis, experimental and CoMFA-based predicted activity for all the compounds, expressed as pIC50 (Table S4), were used to plot the predicted versus the experimental pIC50 values for all the test compounds, indicating good correlation for both training and test datasets (Figure6d). Additionally, this CoMFA model resulted into the electrostatic (35.3%) and steric (29.2%) fields outputs (stdev*coeff), whose translated contour surfaces of the corresponding field contributions, including aligned test compounds, are displayed in Figure6a–b. Molecules 2019, 24, x FOR PEER REVIEW 10 of 19

Figure 6.FigureComparative 6. Comparative molecular molecular field field analysis analysis (CoMFA). (CoMFA). StericSteric ( a)) and and Electrostatic Electrostatic (b) (mapsb) maps after after the CoMFAthe model. CoMFA Green model. and Green yellow and contours yellow contours expose expose positions positions where where a bulky a bulky group group would would be be favorable favorable and unfavorable, respectively. Blue and red contours comprise regions where an increase and unfavorable, respectively. Blue and red contours comprise regions where an increase of positive of positive and negative charges, respectively, will enhance the antifungal activity; (c) schematic and negative charges, respectively, will enhance the antifungal activity; (c) schematic representation representation of the steric and electrostatic contributions according to CoMFA using 24 as model

of the stericcompound; and electrostatic(d) CoMFA predicted contributions as experimental according pIC50 tovalues. CoMFA using 24 as model compound; (d) CoMFA predicted as experimental pIC50 values. 2.5. Binding Mode and Residual Interactions As mentioned above, the behavior of docking scores per enzyme within dataset was found to be different. These differences were found to be depending on the particular enzyme and ligand involved into the docking simulation. Thus, Table S5 exhibited the information of the best-docked phytoalexin analog for each enzyme, which compiles the interacting residues of the respective enzymatic active site and the interacting moieties for the ligands, according to individual two-dimensional residual interaction diagrams (Figure S3). After the analysis of these diagrams to outline some insights into the binding mode of the simulated complexes, several types of interactions were evidenced, involving hydrogen (H)-bonds and hydrophobic contacts, common Van der Waals interactions as well as π-π, π-sulfide, π-cation and π-anion interactions. These interactions were taken as the key contacts and an indicative of the importance of the presence of sulfur (as thione) and aromatic rings (indole and EWG) as crucial moieties for interacting with the respective residues of active site and stabilizing the resulting enzyme-ligand complexes. In this regard, compounds 5 and 24 were the best-docked phytoalexin analogs for the corresponding calculation with enzyme E25. The respective complexes (i.e., E25:5 and E25:24) exhibited the lowest Vina scores into the whole dataset (−11.68 ± 0.19 and −13.25 ± 1.02kcal/mol, respectively), although other good Vina scores were also obtained with these compounds and other enzymes, as previously mentioned (see Figure 3 and Table S2). E25 is a lanosterol 14α-demethylase (LDM or CYP51), an important enzyme target for several human fungal pathogens. It is a cytochrome P450 enzyme that catalyzes the conversion of lanosterol to 4,4-dimethylcholesta-8(9),14,24-trien-3β-ol (through the C-14α demethylation) [24]. This is a critical step in the sterols biosynthesis, because is the early waypoint in the lanosterol conversion to other important sterols to the cell. Selective inhibition of LDM produces lanosterol accumulation as well as other 14-methyl sterols (often toxic to the fungus), causing fungal growth inhibition [24]. Within the arsenal of fungicides, azoles are reported as selective inhibitors of fungal LDM, such as fluconazole, ketoconazole voriconazole, and itraconazole, frequently used for systemic and topical

Molecules 2020, 25, 45 10 of 19

Regarding the electrostatic field contour map (Figure6a), the positive e ffect of positively and negatively-charged regions on antifungal activity can be depicted as blue and red contours, respectively. In the present analysis, electron-deficient substituents were found to be not crucial for the antifungal activity, because any significant blue contour was obtained after the PLS regression. Thus, the activity can be therefore enhanced by the presence of electron-rich groups attached at sulfur in dithiocarbamate moiety, as the electron-withdrawing groups (EWG), and other typical electronegative groups (i.e., heteroatoms). In the case of the contour map of steric field (Figure6b), unfavorable and favorable zones by steric effects were then symbolized as yellow and green contours, respectively. Thus, the antifungal activity can be enhanced if a bulky steric group is located on the alkyl group (as a large n-alkyl substituent) at ester moiety as well as on the EWG, especially in the (3-oxo-1,3-diphenyl)propyl moiety of ODP-DTC-type compounds. In contrast, the antifungal activity is disfavored if a bulky group is placed as a ramification at C2 in the n-alkyl group at ester moiety. These results demonstrated that a bulky group with electron-withdrawing nature at sulfur site in dithiocarbamate moiety (such as a chalcone precursor) can improve the antifungal activity. From these facts, a two-dimensional scheme, involving the respective representations of electrostatic and steric effects of test compounds, is exposed in Figure6c. Accordingly, the resulting model suggested the presence of both bulky and electronegative substituents adequately oriented at the sterically and electrostatically favorable regions, respectively. Both sterically bulky and electron-rich substituents are therefore required together to increase the mycelial growth inhibitory action in test compounds. In this context, compound 24 fitted very well to the above-mentioned structural prerequisites to be a promising antifungal agent. Similar trends were observed from a CoMFA model obtained with close related N,S-dialkyl dithiocarbamates using other aminoacids (i.e., l-alanine, l-phenylalanine, and l-tyrosine) [9] instead l-tryptophan.

2.5. Binding Mode and Residual Interactions As mentioned above, the behavior of docking scores per enzyme within dataset was found to be different. These differences were found to be depending on the particular enzyme and ligand involved into the docking simulation. Thus, Table S5 exhibited the information of the best-docked phytoalexin analog for each enzyme, which compiles the interacting residues of the respective enzymatic active site and the interacting moieties for the ligands, according to individual two-dimensional residual interaction diagrams (Figure S3). After the analysis of these diagrams to outline some insights into the binding mode of the simulated complexes, several types of interactions were evidenced, involving hydrogen (H)-bonds and hydrophobic contacts, common Van der Waals interactions as well as π-π, π-sulfide, π-cation and π-anion interactions. These interactions were taken as the key contacts and an indicative of the importance of the presence of sulfur (as thione) and aromatic rings (indole and EWG) as crucial moieties for interacting with the respective residues of active site and stabilizing the resulting enzyme-ligand complexes. In this regard, compounds 5 and 24 were the best-docked phytoalexin analogs for the corresponding calculation with enzyme E25. The respective complexes (i.e., E25:5 and E25:24) exhibited the lowest Vina scores into the whole dataset ( 11.68 0.19 and − ± 13.25 1.02 kcal/mol, respectively), although other good Vina scores were also obtained with these − ± compounds and other enzymes, as previously mentioned (see Figure3 and Table S2). E25 is a lanosterol 14α-demethylase (LDM or CYP51), an important enzyme target for several human fungal pathogens. It is a cytochrome P450 enzyme that catalyzes the conversion of lanosterol to 4,4-dimethylcholesta-8(9),14,24-trien-3β-ol (through the C-14α demethylation) [24]. This is a critical step in the sterols biosynthesis, because is the early waypoint in the lanosterol conversion to other important sterols to the cell. Selective inhibition of LDM produces lanosterol accumulation as well as other 14-methyl sterols (often toxic to the fungus), causing fungal growth inhibition [24]. Within the arsenal of fungicides, azoles are reported as selective inhibitors of fungal LDM, such as fluconazole, ketoconazole voriconazole, and itraconazole, frequently used for systemic and topical mycoses [25]. However, the examination of the ergosterol depletion in several, different phytopathogens is still a topic to be explored and extended in order to develop selective inhibitors of this fungal vital pathway. Molecules 2019, 24, x FOR PEER REVIEW 11 of 19 mycosesMolecules 2020 [25]., 25, 45However, the examination of the ergosterol depletion in several, different11 of 19 phytopathogens is still a topic to be explored and extended in order to develop selective inhibitors of this fungal vital pathway. In the case of Fusarium spp., the crystal structure of LDM has not been reportedIn the case yet. of Therefore,Fusarium spp.,a homology the crystal model structure was then of LDMbuilt hasusing not the been reported reported enzyme yet. Therefore,sequence of a thehomology transmembrane model was LDM then from built F. usingcerealis the (Uniprot reported entry: enzyme I6ZLS0). sequence The resulting of the transmembrane homology model LDM of thefrom apoenzymeF. cerealis of(Uniprot F. cerealis entry: lanosterol I6ZLS0). 14α-demethylase The resulting (Fc homologyLDM) (Z-score model = 2.191) of the was apoenzyme embedded of inF. acerealis 266-lipidslanosterol bilayer 14α -demethylaseof 1-palmitoyl-2-oleoylph (FcLDM) (Z-scoreosphatidylethanolamine= 2.191) was embedded (OPPEA) in a 266-lipids and SPC bilayer water of molecules,1-palmitoyl-2-oleoylphosphatidylethanolamine and chloride and sodium ions, was (OPPEA) then andoptimized SPC water through molecules, a 5-ns and conventional chloride and molecularsodium ions, dynamics was then (MD) optimized simulation. through Final a 5-nsoptimized conventional model molecularof E25, i.e., dynamics FcLDM (MD)(Figure simulation. 7a), was subsequentlyFinal optimized used model for ofthe E25, above-mentioned i.e., FcLDM (Figure molecular7a), was docking subsequently calculations used for and the further above-mentioned molecular dynamicsmolecular dockingsimulations. calculations Compound and further 24 exhibited molecular the dynamics best Vina simulations. score with Compound this enzyme24 exhibited (as above-mentioned)the best Vina score and with it exhibited this enzyme one (as of above-mentioned)the lowest IC50 value, and indicating it exhibited its promising one of the behavior lowest IC as50 antifungal.value, indicating Molegro its promisingvirtual docking behavior (MVD) as antifungal. was also used Molegro to simulate virtual this docking intermolecular (MVD) was interaction also used asto simulaterescoring this strategy, intermolecular and compound interaction 24 asexhibited rescoring the strategy, best score and compound (MVD score24 exhibited= −167.16). the Both best score (MVD score = 167.16). Both best-docked poses of 24, from Autodock/Vina and MVD, exhibited best-docked poses of− 24, from Autodock/Vina and MVD, exhibited an RMSD below 1, which confirmsan RMSD the below good 1, performance which confirms of our the molecular good performance docking ofresults. our molecular docking results.

FigureFigure 7. ((aa)) F.F. cerealis lanosterol 14 α-demethylase ( (FcLDM), E25, E25, model, embedded in in a 266-lipids bilayerbilayer of of 1-palmitoyl-2-oleoylphosphatidylethanolamine (OPPEA) (OPPEA) and and docked docked with with compound compound 2424; ; ((bb)) Interaction Interaction surface surface of of E25 E25 and and best-docked best-docked pose pose of of 2424 (contour(contour related related to to the the donor donor (pink) (pink) and and acceptoracceptor (green) (green) zones zones with with the the interface); interface); ( (cc)) three-dimensional three-dimensional (3D) (3D) model model of of lowest-energy lowest-energy docked docked pose of of 24 within active site of FcLDM (selected residues residues at at active active site site of of FcFcLDMLDM are are marked marked in in gray gray lightlight sticks; sticks; remaining remaining enzyme enzyme polycolored polycolored cartoon; cartoon; ligand ligand 2424 inin gray gray bold bold sticks; sticks; ( (dd)) 3D 3D model model of of solvent-accessiblesolvent-accessible surface surface area area (SASA) (SASA) of of FcFcLDMLDM with with the the lowest-energy lowest-energy docked pose of of 2424 withinwithin its its activeactive site site (ligand (ligand 2424 inin yellow yellow sticks). sticks).

ThisThis compound compound fitted fitted adequately adequately within within the the active active site site of of FcFcLDMLDM (Figure (Figure 77d),d), locatedlocated overover thethe membranemembrane forfor thethe catalytic catalytic action action of this of transmembranethis transmembrane enzyme enzyme (Figure 7a).(Figure The H7a).-bond The interacting H-bond interactingsurface of thesurface enzyme of the (Figure enzyme7b) (Figure indicated 7b)that indicatedFcLDM that residues FcLDM canresidues be oriented can be tooriented have a to role have as aacceptor, role as acceptor, but the complex but the complex exhibited exhibited important important non-polar non-polar interactions interactions with Tyr126, with Tyr126, Phe241, Phe241, Phe286, Phe286,His381 andHis381 Phe384 and residues Phe384 (Figureresidues7c). (Figure The respective 7c). The two-dimensionalrespective two-dime (2D)-residualnsional (2D)-residual interaction interactiondiagram (Figure diagram8a) indicated(Figure 8a) that indicated these key that contacts these key are relatedcontacts to areπ-sulfur related (with to π-sulfur His381 (with and Phe241), His381 andπ-π stackedPhe241), (with π-π stacked Tyr126 and(with Phe236), Tyr126 andπ-π T-shapedPhe236), π (with-π T-shaped Phe384) (with interaction Phe384) types. interaction Heme types. group did not interact with compound 24, although it is located close to a phenyl group, which implies

Molecules 2019, 24, x FOR PEER REVIEW 12 of 19 Molecules 2020, 25, 45 12 of 19 Heme group did not interact with compound 24, although it is located close to a phenyl group, which implies an advantage in terms of a plausible transformation of the test ligand. In this sense, a ansuccessful advantage interaction in terms can of abe plausible produced, transformation avoiding a posterior of the test catalytic ligand. conversion In this sense, by athis successful HEME interactiongroup. The can ligand be produced, interaction avoiding plot (Figure a posterior 8b) confir catalyticms the conversion previous byinformation, this HEME indicating group. The Phe241, ligand interactionPhe236, Phe384 plot (Figure and Tyr1268b) confirms as crucial the previousresidues information,to interact in indicating order to Phe241,stabilize Phe236, the ligand-enzyme Phe384 and Tyr126complex. as crucialAccording residues to the to above interact analysis, in order it to is stabilize possible the to conclude ligand-enzyme that the complex. E25 is a According plausible totarget the aboveto be inhibited analysis, itby is 24 possible, due toto its conclude satisfactory that in the silico E25 interaction is a plausible profile target with to be this inhibited phytoalexin by 24 ,analog. due to itsHowever, satisfactory this in information silico interaction had profileto be extended with this phytoalexincomputationally analog. through However, a molecular this information dynamics had tosimulation. be extended computationally through a molecular dynamics simulation.

FigureFigure 8.8.Two-dimensional Two-dimensional (2D) (2D) plots plots for FcforLDM FcLDM (E25): 24(E25):complex.24 complex. (a) 2D-residual (a) 2D-residual interaction interaction diagram usingdiagram Discovery using Discovery Studio Client Studio (interaction Client (interaction type is diff erentiatedtype is differentiated by colored circlesby colored (residues) circles and (residues) dashed linesand dashed (directed lines to the (directed ligand moiety)to the ligand according moiety) to the according convention); to the (b convention);) ligand interaction (b) ligand diagram interaction (LID) usingdiagram Maestro (LID) 11.5using [light Maestro green 11.5 lines [light depicts green the lines active de sitepicts surface, the active dark site green surface, lines dark indicate green the linesπ-π stackingindicate betweenthe π-π stacking residue andbetween aromatic residue moieties and aromatic of 24; residues moieties are of di 24fferentiated; residues are by colorsdifferentiated according by tocolors the interactionaccording typeto the as indicatedinteraction in type LID legend:as indicated hydrophobic in LID (green),legend: polarhydrophobic (aquamarine), (green), positive polar charge(aquamarine), (purple)]. positive charge (purple)]. 2.6. Molecular Dynamics Simulations 2.6. Molecular Dynamics Simulations Lanosterol 14α-demethylase (LDM) from F. cerealis was selected as test enzyme for 90-ns molecular Lanosterol 14α-demethylase (LDM) from F. cerealis was selected as test enzyme for 90-ns dynamics simulations due to its better molecular docking behavior with the test phytoalexin analogs, molecular dynamics simulations due to its better molecular docking behavior with the test specifically compound 24. The 3D structure of this enzyme was built through homology modeling phytoalexin analogs, specifically compound 24. The 3D structure of this enzyme was built through using Yasara Structure [26], since no crystal data is available, embedded in a 266-lipids OPPEA bilayer. homology modeling using Yasara Structure [26], since no crystal data is available, embedded in a This simulation was implemented as the next step to expand the information about the binding mode 266-lipids OPPEA bilayer. This simulation was implemented as the next step to expand the of this promising candidate. information about the binding mode of this promising candidate. Ligand:protein trajectories for those complexes between 24 and FcLDM were monitored by means Ligand:protein trajectories for those complexes between 24 and FcLDM were monitored by of geometric properties over the time. Thus, root mean square deviations (RMSD) were used to evaluate means of geometric properties over the time. Thus, root mean square deviations (RMSD) were used the structural stability of the receptor frame by measuring the time-dependent distance between to evaluate the structural stability of the receptor frame by measuring the time-dependent distance different positions (in Å) of the set of atoms (Figure9a). For this, separately 90-ns molecular dynamics between different positions (in Å) of the set of atoms (Figure 9a). For this, separately 90-ns molecular simulations for the FcLDM (E25) alone (apoenzyme) and docked distinctly with 24, lanosterol (Lan) dynamics simulations for the FcLDM (E25) alone (apoenzyme) and docked distinctly with 24, and brassinin (Bra) were recorded for comparing purposes. As result, the apoenzyme exhibited a lanosterol (Lan) and brassinin (Bra) were recorded for comparing purposes. As result, the normal evolution during the initial part of the simulation but revealed a slight perturbation at 20 ns and apoenzyme exhibited a normal evolution during the initial part of the simulation but revealed a then a good stabilization over the remaining simulation. Lanosterol avoid such a perturbation and the slight perturbation at 20 ns and then a good stabilization over the remaining simulation. Lanosterol complex FcLDM:lanosterol reached stability at 15 ns, whereas brassinin displayed the most-perturbed avoid such a perturbation and the complex FcLDM:lanosterol reached stability at 15 ns, whereas profile. However, the FcLDM:24 complex achieved early the stability and it is maintained over entire brassinin displayed the most-perturbed profile. However, the FcLDM:24 complex achieved early the simulation. In addition, the fluctuations of the atomic positions for each residue of the enzyme, the stability and it is maintained over entire simulation. In addition, the fluctuations of the atomic root mean square fluctuation (RMSF) were also considered to scrutinize the flexibility and secondary positions for each residue of the enzyme, the root mean square fluctuation (RMSF) were also structure of the FcLDM enzyme under interaction with the above-mentioned test ligands. All examined

Molecules 2019, 24, x FOR PEER REVIEW 13 of 19 Molecules 2020, 25, 45 13 of 19 considered to scrutinize the flexibility and secondary structure of the FcLDM enzyme under interaction with the above-mentioned test ligands. All examined complexes showed different complexesbehavior (Figure showed 9b), diff erentwith behaviorexplicit differences (Figure9b), in with some explicit regions, di ff erencesfluctuating in some significantly regions, fluctuating along the significantlyMD simulations along trajectory the MD simulations between 0.2 trajectory and 1.0 between nm. In 0.2this and regard, 1.0 nm. test In thisligands regard, (i.e., test brassinin, ligands (i.e.,lanosterol brassinin, and lanosterol 24) exhibited and 24 )different exhibited RMSF different profiles, RMSF profiles,specifically specifically through through fluctuations fluctuations by the by theparticular particular interacting interacting residues, residues, but but a acommon common fluctuating fluctuating zone zone was was observed observed (Arg375–Asp385). InIn thethe specificspecific case case of Fcof LDM:FcLDM:24 complex,24 complex, fluctuations fluctuations were foundwere tofound be around to be thearound previously the previously identified crucialidentified contacts crucial (i.e., contacts Tyr126, (i.e., Phe241, Tyr126, Phe286, Phe241, His381 Phe286, and His381 Phe384), and butPhe384), the association but the association of 24 did of not 24 obviouslydid not obviously affect the flexibilityaffect the offlex thoseibility important of those fragments important to fragments maintain overall to maintain system overall stability system (such residuesstability with(such RMSF residues> 0.2 with nm) inRMSF comparison > 0.2 nm) to the inapo comparisonLDM; thus, to the the inhibition apoLDM; mode thus, may the be inhibition achieved bymode stabilizing may be the achieved system. by These stabilizing results indicatedthe system. the These excellent results in silico indicated interaction the excellent profile of in24 silicowith theinteractionFcLDM enzymeprofile of over 24 with the time.the FcLDM enzyme over the time.

FigureFigure 9.9. ((a)) RootRoot meanmean squaresquare deviationsdeviations (RMSD),(RMSD), alongalong thethe trajectoriestrajectories simulatedsimulated byby molecularmolecular dynamicsdynamics duringduring 9090 ns,ns, forfor thethe FcFcLDMLDM (E25)(E25) (apoenzyme),(apoenzyme), andand dockeddocked separatelyseparately withwith 24,24, lanosterollanosterol (Lan)(Lan) andand brassinin brassinin (Bra); (Bra); (b) root(b) meanroot mean quadratic quadratic fluctuations fluctuations (RMSF) along(RMSF) the along trajectories the trajectories simulated bysimulated molecular bydynamics molecular during dynamics 90 ns, during for the 90Fc LDMns, for (E25) the (apoenzyme),FcLDM (E25) and (apoenzyme), docked separately and docked with 24,separatelylanosterol with (Lan) 24, andlanosterol brassinin (Lan) (Bra). and brassinin (Bra). 2.7. Binding Free Energy 2.7. Binding Free Energy The binding free energy for compound 24, lanosterol and brassinin during interaction of The binding free energy for compound 24, lanosterol and brassinin during interaction of FcLDM for the last 20 ns of MD trajectory was estimated by MM/PBSA approach, using the MD FcLDM for the last 20 ns of MD trajectory was estimated by MM/PBSA approach, using the MD simulations. All three ligands exhibited negative binding energies, but compound 24 showed lower simulations. All three ligands exhibited negative binding energies, but compound 24 showed lower binding energy to that of brassinin ( 121.6 7.1 kJ/mol versus 39.7 5.4 kJ/mol, respectively) and binding energy to that of brassinin −(−121.6 ±± 7.1 kJ/mol versus −−39.7 ±± 5.4 kJ/mol, respectively) and lanosterol ( 101.3 6.5 kJ/mol), which rationalize the observed best docking performance and the lanosterol (−−101.3 ± 6.5 kJ/mol), which rationalize the observed best docking performance and the best experimental antifungal activity of 24. The main contribution to the binding energy was due best experimental antifungal activity of 24. The main contribution to the binding energy was due to to vdW energies (> 90 kJ/mol). This fact was then confirmed after binding energy decomposition vdW energies (> −90− kJ/mol). This fact was then confirmed after binding energy decomposition on on the residues of FcLDM:24 complex, since the most binding energy contributing residues were the residues of FcLDM:24 complex, since the most binding energy contributing residues were found found to be Phe241 ( 9.2 kJ/mol), Phe384 ( 4.3 kJ/mol), Phe236 ( 3.5 kJ/mol), His381 ( 3.2 kJ/mol) and to be Phe241 (−9.2 −kJ/mol), Phe384 (−4.3− kJ/mol), Phe236 (−3.5− kJ/mol), His381 (−−3.2 kJ/mol) and Tyr126 ( 2.8 kJ/mol). These results suggested that non-polar electrostatic interactions are the main Tyr126 (−−2.8 kJ/mol). These results suggested that non-polar electrostatic interactions are the main driving force for molecular recognition of FcLDM by 24. However, considering the MD simulations driving force for molecular recognition of FcLDM by 24. However, considering the MD simulations and binding affinity calculations, the resulting simulated data should be contemplated carefully, since and binding affinity calculations, the resulting simulated data should be contemplated carefully, they were extracted from a single trajectory analysis and the convergence could be hardly obtained for since they were extracted from a single trajectory analysis and the convergence could be hardly a highly fluctuating system. obtained for a highly fluctuating system. 3. Materials and Methods 3. Materials and Methods 3.1. Design and Synthesis of Indole-Containing Phytoalexin Analogs 3.1. Design and Synthesis of Indole-Containing Phytoalexin Analogs Twenty-one indole-containing phytoalexin analogs and four L-tryptophan alkyl esters (1–25) were synthesized through the previously reported protocol [8] (Figure S1) and they were chosen as test

Molecules 2020, 25, 45 14 of 19 compounds (Figure1). These structures were 3D-sketched and organized according to close-related moieties affording seven substituted groups such as 2-alkyl aminoesters (1–4), N,N-dialkylthioureas (5–8), 2-cyanoethyl N-alkyldithiocarbamates (9–12), 3-methoxy-3-oxopropyl N-alkyldithiocarbamate (13–16), 2-methyl-4-oxopentan-2-yl N-alkyldithiocarbamate (17–20), 2-oxo-1,3-diphenylpropyl N-alkyldithiocarbamates (21–24) and 4-[(1H-indol-3-yl)-methylene]-2-sulfanylidene-1,3-thiazolidin-5-one (25). Brassinin (26) was included into the study and used as reference compound. A previous MMFF geometry minimization for each ligand was carried out in Spartan’14 (Wavefunction, Irvine, CA, USA) without any geometrical restrictions, prior to the docking and dynamics simulations.

3.2. Enzymes Twenty-five fungal enzymes were searched and selected due to their important implications for Fusarium spp. in metabolic processes or invasive properties (Table A1). The crystallographic structure of twenty of these enzymes with an inhibitor or substrate was retrieved from the RCBS Protein Data Bank (PDB) (http://www.rcsb.org/). The three-dimensional (3D) structures of the remaining five enzymes were obtained by homology modeling using the standard protocol included into the Yasara Structure package (http://www.yasara.org/)[26]. Water molecules, ligands and other heteroatoms were removed from the protein molecule. Hydrogen atoms were added to the protein. These crystallographic structures were retained without any processing for molecular docking. The co-crystallized inhibitors/substrates were thus employed to define the corresponding active site, the flexible residues within this active site and as a validation criterion of docking calculations (re-docking). PDB-files were then effectively prepared using the AutoDock Vina plugin under PyMOL, and saved as pdbqt-files. The structural information of the employed enzymes is shown in Table A1.

3.3. Molecular Docking Molecular docking was carried out using Autodock/Vina (Molecular Graphics Lab, The Scripps Research Institute, La Jolla, CA, USA) using the AMBER force field with PyMOL 2.3 (Schrödinger LLC, New York, NY, USA) as molecular graphics system. The protein and ligand molecules were prepared as described above. The active site for each enzyme was located into a grid and the flexible residues were selected according to those interacting with co-crystallized inhibitors or substrates at 4.0 Å. All molecular docking calculations were carried out using flexible residues and performed between the optimized ligand into the active site through a cube at the geometrical center of the native ligand present in the evaluated PDB structure, with the dimensions 24 24 24 Å, covering × × the ligand binding site with a grid point spacing of 0.375 Å. Each calculation involved ten replicates starting from separated pdb-files of each enzyme and ligand to evaluate convergence. Molecular docking for the co-crystallized inhibitors/substrates was also carried out as a reference to evaluate the docking performance through re-docking. Additionally, the specificity and sensitivity of the docking protocol, 25 compounds of diverse chemical nature, having IC50 < 5 µM, against five test enzymes (isocitrate lyase, aspartate kinase, sereine esterase, lanosterol 14α-demethylase and trichothecene acetyltransferase) were compiled from the literature and available databases such as ChEMBL [27] and pubchem [28]. Forty-five decoys per each active compound were compiled from the directory of useful decoys (DUD-E) [29] and CheMBL databases. Thus, decoys and bioactives were processed using the docking process. The resulting data was then assessed through a receiver operating characteristic (ROC) and scores enrichment using the screening explorer webserver [30]. Once the docking parameters were validated, the docking simulations of test indole-containing phytoalexin analogs were performed. The resulting binding modes were firstly analyzed in PyMOL 2.3 and the docking poses were ranked according to their docking Vina scores (as affinities in kcal/mol). Molegro Virtual Docker 6.0 (CLC Bio Company, Aarhus, Denmark) was also used to evaluate secondly the best-poses of best-docked compounds as a rescoring strategy. Both the ranked list of docked ligands and their corresponding binding poses were exported as a CSV file for further analysis. Moreover, two-dimensional (2D) residual interactions diagrams and 3D figures of these enzymes were obtained using Discovery Studio Molecules 2020, 25, 45 15 of 19

Client 16.1 (Biovia, San Diego, CA, USA) and Maestro Elements 8.3 (Schrödinger, Cambridge, MA, USA) from the Vina outputs (pdbqt-files) for compounds that exhibited highest affinity. All calculations were performed in a Dual Intel Xeon® processor CPU @ 2.6 GHz of Intel system origin, with 64 GB DDR3 RAM on an Ubuntu 12.04 server.

3.4. Statistical Analysis Descriptive and inferential statistics tests were carried out such as principal component analysis, significance tests and partial least-square (PLS) regression. Orthogonal partial least squares-discriminant analysis (OPLS-DA) and single-Y OPLS were carried out in SIMCA 14.0 software (Umetrics, Umeå, Sweden) using the docking affinity values and the reported IC50 values in order to observe feasible relationships.

3.5. Molecular Dynamics Simulations of F. Cerealis Lanosterol 14α-demethylase (FcLDM) The resulting homology model of the apoenzyme of F. cerealis lanosterol 14α-demethylase (FcLDM) were embedded in a 1-palmitoyl-2-oleoylphosphatidylethanolamine (OPPEA) bilayer of 266 lipids covering an area of 125 125 Å2, using a cell box with size of 125 125 200 Å3 filled with SPC water × × × molecules for solvation. The net protein charge was neutralized by 60 chloride and 54 sodium ions. This membrane-FcLDM model was built using the standard protocol included into the Yasara Structure package [26]. The most plausible model was identified through steepest descent method-based energy minimization until 100 kJ/(mol nm) energy convergence, followed by a 250-ps restrained · equilibration simulation and a 5-ns conventional molecular dynamics (MD) simulation. The same protocol was carried out on the simulation systems of FcLDM+lanosterol, FcLDM+brassinin and FcLDM + 24 (obtained after molecular docking), which were individually subjected to an additional 90-ns conventional MD simulation for exploring the performance of these ligands within the active site of the enzyme. These MD simulations were run in Gromacs 5.0.5 (open source, http://www.gromacs.org) on an Ubuntu 12.04 server, using NPT and periodic boundary conditions, as previously reported [31]. Hence, docked ligands were prepared by adding hydrogen atoms in UCSF Chimera (UCSF, CA, USA) and the resulting pdb-file was uploaded to the atb server (http://compbio.biosci.uq.edu.au/atb/) to add the respective Gromo53a6 force field and get the itp-type topology file. Membrane-FcLDM topologies were obtained in Gromacs using the Gromos53a6 force fields, owing to the presence of Heme group. The SPC water model was then implemented for solvation, in a triclinic box using a 1.0-nm margin distance. 0.10 M NaCl was added to the simulation systems and water molecules were randomly replaced until neutrality. An energy minimization through 2000-steps steepest descent method was then used. NVT equilibration at 310 K during 500 ps, followed by NPT equilibration during 1000 ps using the Parrinello-Rahman method at 1 bar as reference were done on the systems using position restraints. Finally, solute position restraints were released and a production run for 90 ns was performed. Temperature and pressure were kept constant at 310 K and 1 bar, respectively. Coordinates were recorded in a 1 fs time step. Electrostatic forces were calculated using the particle-mesh Ewald (PME) method. Periodic boundary conditions were used in all simulations and covalent bond lengths were constrained by the LINCS algorithm.

3.6. Binding Free Energy Analyses Binding free energy was calculated using the g_mmpbsa tool [32](open source drug discovery consortium (OSDD), New Dehli, India). This tool calculated components of the free energy of the protein–substrate binding (∆GBind), using the molecular mechanic/poisson-boltzmann surface area (MM/PBSA) method [33,34]. In this method, ∆Gbind calculation between a protein and a ligand is carried out by ∆Gbind = ∆H T∆S ∆EMM + ∆Gsol T∆S, ∆EMM = ∆Einternal + ∆Eelectrostatic ∆EvdW, elec vdW− ≈ − − ∆Gsolv = ∆G solv + DG solv. Here, the total gas phase energy on the binding of MM energy is ∆EMM, the free energy of solvation is ∆Gsolv and the entropy contribution is T∆S. Poisson-Boltzmann model was used to compute the electrostatic solvation energy in a continuum solvent. The derivation Molecules 2020, 25, 45 16 of 19

of non-polar solvation energy term was computed as solvent-accessible surface area (SASA). ∆EMM were calculated using the Lennard-Jones and Coulomb potential [33,34]. ∆GBind was used to analyze the binding associations between FcLDM and selected ligands (i.e., lanosterol, 24 and brassinin) by decomposing the total binding free energy into each residue. The binding energy calculations of the selected ligands were performed for 100 snapshots taken at an interval of 200 ps during the last sTable 20-ns MD simulations.

3.7. Antifungal Assay Compounds 1–25 were evaluated for their antifungal activity against F. oxysporum following the previously reported amended-medium protocol to assess the in-vitro mycelial growth inhibition [8]. The activity was expressed as half-maximal inhibitory concentrations, IC50 (in mM), obtained after non-linear regression using the program GraphPad Prism version 5.00 (GraphPad Software, San Diego, CA, USA).

3.8. Comparative Molecular Field Analysis (CoMFA) Best-docked poses of test ligands were merged and aligned by means of particular tethers placed on indole, sulfur and aromatic rings moieties, using the molecular overlay tool included in the software Discovery Studio Client 16.1 (Biovia, San Diego, CA, USA). The best-scored compound (i.e., 24) was selected as template to select tethers. The aligned molecules set (in a sdf-file) was randomly divided into two subsets (training and test sets, corresponding to 70 and 30%, respectively), and CoMFA analysis was then performed using Open3DQSAR, using the standard protocol [35]. The experimental antifungal activity evaluated as F. oxysporum mycelial growth inhibition (expressed as IC50 in M) were was converted into negative logarithmic form (pIC50) and then used as independent variable. The models were validated by leave-one-out (LOO) and leave-many-out (LMO) methods. The quality of the models after validation was evaluated, predicting the independent variable for the test set.

4. Conclusions Twenty-one indole-containing phytoalexin analogs and four L-tryptophan alkyl esters (1–25) were docked within the active site of a set of 25 enzymes of Fusarium spp. through Autodock/Vina. NNDATU, ODP-TDC and IST-type analogs exhibited the best docking scores and interaction profile with several enzymes, especially nitroalkane oxidases (E17 and E18) and FMN-dependent dehydrogenase (E24) and 14-lanosterol demethylase (E25), respectively, according to the distribution and correlation analyses as well as discriminant analysis by OPLS. Computational results were also integrated to experimental in vitro (IC50) mycelial growth inhibition against F. oxysporum, exhibiting good correlations between docking scores and antifungal data (by single-Y OPLS) as well as among structures and antifungal data (on the basis of a robust CoMFA model). The CoMFA-derived PLS-regression indicated that sterically bulky and electron-rich substituents are jointly required to increase the mycelial growth inhibitory action in test compounds, indicating ODP-TD-type compounds can be considered as candidates for phytoalexin bioisosteres. E25 (a homology model of a F. cerealis 14-lanosterol demethylase) was taken as study model owing to the well-behaved interaction with compound 24 (tert-butyl (((3-oxo-1,3-diphenylpropyl)thio)carbonothioyl)-L-tryptophanate). The binding mode was examined, and some residues were found to be important to stabilize the FcLDM:24 complex, which was extended over the time by 90-ns molecular dynamics simulations. In this regard, the non-polar electrostatic interactions can be considered as the main driving force for the molecular recognition of FcLDM by 24. From all of the obtained results, compound 24 can be considered as a putative inhibitor of this enzyme, rationalizing the good experimental antifungal activity through in vitro mycelial growth inhibition (IC50 = 0.44 mM). Such a mode of action will be further explored by in vitro enzymatic inhibition over purified enzyme, which is currently ongoing. Thus, after data analyses, some important hits, putative enzyme targets and structural requirements were established, indicating the potential of indole-containing phytoalexin-based bioisosteres as an alternative for the control of Fusarium spp.-like Molecules 2020, 25, 45 17 of 19 phytopathogens. Hence, this information will be preserved during our further studies focused on the development of antifungals based on phytoalexin analogs.

Supplementary Materials: The following are available online at http://www.mdpi.com/1420-3049/25/1/45/s1, Table S1: IUPAC names of test phytoalexin analogues. Table S2: Calculated Vina scores for compounds 1–26 docked with enzymes E1–E25. Table S3: Values for the Pearson correlation of the enzymes set. Table S4: Antifungal activity against Fusarium oxysporum through amended-medium assay. Table S5: Recorded interactions according to the best-docked pose of test phytoalexin analogs. Figure S1: Reaction route for obtaining the test indole-containing phytoalexin analogs. Figure S2: Principal component analysis (PCA)-derived score plot. Unsupervised multivariate analysis from affinity dataset. Figure S3: 2D-residual interactions for the test indole-containing phytoalexin analogs. Author Contributions: Conceptualization, D.Q. and E.C.-B.; methodology, A.A.-R., D.Q. and E.C.-B.; software, A.A.-R. and E.C.-B.; validation, A.A.-R. and E.C.-B.; formal analysis, A.A.-R., D.Q. and E.C.-B.; investigation, A.A.-R., D.Q. and E.C.-B.; resources, D.Q. and E.C.-B.; data curation, E.C.-B.; writing—original draft preparation, A.A.-R.; writing—review and editing, A.A.-R., D.Q. and E.C.-B.; project administration, D.Q. funding acquisition, D.Q. and E.C.-B. All authors have read and agreed to the published version of the manuscript. Funding: The present work is a product derived by the Project INV-CIAS-1787 financed by Vicerrectoría de Investigaciones at UMNG—Validity 2015. Acknowledgments: Authors thank UM Nueva Granada for the financial support. Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A

Table A1. Data of test enzymes.

ID Enzyme (Type) PDB Code Source E1 endoglucanase I (glycosyl hydrolase) [NAG cavity] a 1OVW F. oxysporum E2 endoglucanase I (glycosyl hydrolase) [native binding site cavity] a 1OVW F. oxysporum E3 endoglucanase I (glycosyl hydrolase) 4OVW F. oxysporum E4 endoglucanase I (cellobiohydrolase) 3OVW F. oxysporum E5 acetyltransferase (transferase) 2ZBA F. sporotrichioides E6 thrichodiene synthase (synthase) 2PS8 F. sporotrichioides E7 endoglucanase I (cellobiohydrolase) 2OVW F. oxysporum E8 trichodiene synthase (synthase) 1YYQ F. sporotrichioides E9 serine esterase (cutinase) 1XZM F. solani subsp. pisi E10 serine esterase (hydrolase-cutinase) 1XZL F. solani subsp. pisi E11 serine esterase (cutinase) 1OXM F. solani subsp. pisi E12 isocitrate lyase (lyase) 5E9G F. sporotrichioides E13 trichothecene 15-O-acetyltransferase (transferase) 3FP0 F. sporotrichioides E14 isocitrate lyase (lyase) 5E9H F. graminearum E15 trichodiene synthase (synthase) 1JFG F. oxysporum E16 trichothecene 3-O-acetyltransferase (transferase) 3B30 F. graminearum E17 nitroalkane oxidase (NAO) (oxidoreductase) 2REH F. oxysporum E18 nitroalkane oxidase (NAO) (oxidoreductase) 2C0U F. oxysporum E19 flavoenzyme nitroalkane oxidase (oxidoreductase) 3D9E F. oxysporum E20 nitric oxide reductase cytochrome (oxidoreductase) 1ULW F. oxysporum F. vascular E21 NAD(P)-dependent dehydrogenase (dehydrogenase) 1U3T b (A0A0D2YG03 c) F. vascular wilt E22 nitroreductase [NADPH] (oxidoreductase) 2BII b (P39863 c) F. verticilloides E23 aspartate kinase (kinase) 2CDQ b (W7MS01 c) F. verticilloides E24 FMN-dependent dehydrogenase (oxidoreductase) 1GOX b (W7NCP1 c) E25 14-lanosterol demethylase (CYP51) (oxidoreductase) 4LXJ b F. cerealis (I6ZLS0 c) a Two different cavities (binding pockets) evaluated of the same enzyme structure; b Enzymes taken as template (after sequence alignment through homology modeling) from respective Uniprot sequences of Fusarium enzymes using Yasara Structure package [9]; c Uniprot entries of enzyme sequences. Molecules 2020, 25, 45 18 of 19

References

1. Munkvold, G.P. Fusarium species and their associated mycotoxins. In Methods in Molecular Biology; Springer: New York, NY, USA, 2016; pp. 51–106. ISBN 1064-3745. 2. Ma, L.-J.; Geiser, D.M.; Proctor, R.H.; Rooney, A.P.; O’Donnell, K.; Trail, F.; Gardiner, D.M.; Manners, J.M.; Kazan, K. Fusarium pathogenomics. Annu. Rev. Microbiol. 2013, 67, 399–416. [CrossRef] 3. Bashir, M.R.; Atiq, M.; Sajid, M.; Mohsan, M.; Abbas, W.; Alam, M.W.; Bashair, M. Antifungal exploitation of fungicides against Fusarium oxysporum f. sp. capsici causing Fusarium wilt of chilli pepper in Pakistan. Environ. Sci. Pollut. Res. 2017, 25, 6797–6801. [CrossRef] 4. Khomutov, R.M.; Khurs, E.N.; Osipova, T.I.; Zhemchuzhina, N.S.; Mikituk, O.D.; Nazarova, T.A.; Shcherbakova, L.A.; Dzhavakhiya, V.G. Chemical control of adaptive function of plant pathogenic fungi. Dokl. Biochem. Biophys. 2015, 461, 69–71. [CrossRef] 5. Ahuja, I.; Kissen, R.; Bones, A.M. Phytoalexins in defense against pathogens. Trends Plant Sci. 2012, 17, 73–90. [CrossRef][PubMed] 6. Pedras, M.S.C.; Minic, Z.; Jha, M. Brassinin oxidase, a fungal detoxifying enzyme to overcome a plant defense - purification, characterization and inhibition. FEBS J. 2008, 275, 3691–3705. [CrossRef][PubMed] 7. Lima, L.; Barreiro, E. Bioisosterism: A useful strategy for molecular modification and drug design. Curr. Med. Chem. 2005, 12, 23–49. [CrossRef][PubMed] 8. Quiroga, D.; Becerra, L.; Sadat-Bernal, J.; Vargas, N.; Coy-Barrera, E. Synthesis and antifungal activity against Fusarium oxysporum of some Brassinin analogs derived from L-tryptophan: A DFT/B3LYP study on the reaction mechanism. Molecules 2016, 21, 1349. [CrossRef][PubMed] 9. Quiroga, D.; Becerra, L.D.; Coy-Barrera, E. Ultrasound-assisted synthesis, antifungal activity against Fusarium oxysporum, and three-dimensional quantitative structure–activity relationship of N,S-dialkyl dithiocarbamates derived from 2-amino acids. ACS Omega 2019, 4, 13710–13720. [CrossRef][PubMed] 10. Sliwoski, G.; Kothiwale, S.; Meiler, J.; Lowe, E.W., Jr. Computational methods in drug discovery. Pharmacol. Rev. 2013, 66, 334–395. [CrossRef][PubMed] 11. Morris, G.M.; Lim-Wilby, M. Molecular Docking BT—Molecular Modeling of Proteins. In Molecular Modeling of Proteins; Kukol, A., Ed.; Humana Press: Totowa, NJ, USA, 2008; pp. 365–382. ISBN 978-1-59745-177-2. 12. Chen, Y.-C. Beware of docking! Trends Pharmacol. Sci. 2015, 36, 78–95. [CrossRef] 13. Westermaier, Y.; Barril, X.; Scapozza, L. Virtual screening: An in silico tool for interlacing the chemical universe with the proteome. Methods 2015, 71, 44–57. [CrossRef][PubMed] 14. Ma, H.; Bandos, A.I.; Gur, D. On the use of partial area under the ROC curve for comparison of two diagnostic tests. Biom. J. 2015, 57, 304–320. [CrossRef][PubMed] 15. Zhan, D.; Wang, D.; Min, W.; Han, W. Exploring the molecular basis for selective binding of homoserine dehydrogenase from Mycobacterium leprae TN toward inhibitors: A virtual screening study. Int. J. Mol. Sci. 2014, 15, 1826–1841. [CrossRef][PubMed] 16. Garvey, G.S.; Rayment, I.; McCormick, S.P.; Alexander, N.J. Structural and functional characterization of TRI3 trichothecene 15-O-acetyltransferase from Fusarium sporotrichioides. Protein Sci. 2009, 18, 747–761. 17. González, G.; Yutronic, N.; Jara, M. Conformational equilibria in N,N-dialkyltthioureas. Spectrochim. Acta Part A Mol. Spectrosc. 1990, 46, 1729–1736. [CrossRef] 18. Veignie, E.; Ceballos, C.; Len, C.; Rafin, C. Design of new antifungal dithiocarbamic esters having bio-based acrylate moiety. ACS Omega 2019, 4, 4779–4784. [CrossRef] 19. Russell, P.E. A century of fungicide evolution. J. Agric. Sci. 2005, 143, 11–25. [CrossRef] 20. Foster, A.J.; Thines, E. Identification of fungicide targets in pathogenic fungi. In Physiology and Genetics; Springer: Berlin/Heidelberg, Germany, 2009; pp. 233–245. ISBN 9783642002854. 21. Roy, K.; Kar, S.; Das, R.N. Introduction to 3D-QSAR. In Understanding the Basics of QSAR for Applications in Pharmaceutical Sciences and Risk Assessment; Elsevier: London, UK, 2015; pp. 291–317. ISBN 9780128015056. 22. Doweyko, A.M. Three-dimensional quantitative structure–activity relationship: The state of the art. In Comprehensive Medicinal Chemistry II; Elsevier: London, UK, 2007; pp. 575–595. ISBN 9780080450445. 23. Clark, R.D.; Fox, P.C. Statistical variation in progressive scrambling. J. Comput. Aided Mol. Des. 2004, 18, 563–576. [CrossRef] Molecules 2020, 25, 45 19 of 19

24. Sheng, C.; Miao, Z.; Ji, H.; Yao, J.; Wang, W.; Che, X.; Dong, G.; Lu, J.; Guo, W.; Zhang, W. Three-dimensional model of lanosterol 14-demethylase from Cryptococcus neoformans: Active-site characterization and insights into azole binding. Antimicrob. Agents Chemother. 2009, 53, 3487–3495. [CrossRef] 25. Warrilow, A.G.; Parker, J.E.; Kelly, D.E.; Kelly, S.L. Azole affinity of sterol 14α-demethylase (CYP51) enzymes from Candida albicans and Homo sapiens. Antimicrob. Agents Chemother. 2012, 57, 1352–1360. [CrossRef] 26. Land, H.; Humble, M.S. YASARA: A tool to obtain structural guidance in biocatalytic investigations. In Methods in Molecular Biology; Springer: New York, NY, USA, 2017; pp. 43–67. ISBN 1064-3745. 27. Gaulton, A.; Hersey, A.; Nowotka, M.; Bento, A.P.; Chambers, J.; Mendez, D.; Mutowo, P.; Atkinson, F.; Bellis, L.J.; Cibrián-Uhalte, E.; et al. The ChEMBL database in 2017. Nucleic Acids Res. 2017, 45, D945–D954. [CrossRef][PubMed] 28. Kim, S. Getting the most out of PubChem for virtual screening. Expert Opin. Drug Discov. 2016, 11, 843–855. [CrossRef][PubMed] 29. Mysinger, M.M.; Carchia, M.; Irwin, J.J.; Shoichet, B.K. Directory of useful decoys, enhanced (DUD-E): Better ligands and decoys for better benchmarking. J. Med. Chem. 2012, 55, 6582–6594. [CrossRef][PubMed] 30. Empereur-Mot, C.; Zagury, J.-F.; Montes, M. Screening Explorer–an interactive tool for the analysis of screening results. J. Chem. Inf. Model. 2016, 56, 2281–2286. [CrossRef][PubMed] 31. Guerrero-Perilla, C.; Bernal, F.A.; Coy-Barrera, E. Insights into the interaction and binding mode of a set of antifungal azoles as inhibitors of potential fungal enzyme-based targets. Mol. Divers. 2018, 22, 929–942. [CrossRef][PubMed] 32. Kumari, R.; Kumar, R.; Lynn, A. g_mmpbsa—A GROMACS tool for high-throughput MM-PBSA calculations. J. Chem. Inf. Model. 2014, 54, 1951–1962. [CrossRef][PubMed] 33. Sun, H.; Li, Y.; Tian, S.; Xu, L.; Hou, T. Assessing the performance of MM/PBSA and MM/GBSA methods. 4. Accuracies of MM/PBSA and MM/GBSA methodologies evaluated by various simulation protocols using PDBbind data set. Phys. Chem. Chem. Phys. 2014, 16, 16719–16729. [CrossRef] 34. Sun, H.; Li, Y.; Shen, M.; Tian, S.; Xu, L.; Pan, P.; Guan, Y.; Hou, T. Assessing the performance of MM/PBSA and MM/GBSA methods. 5. Improved docking performance using high solute dielectric constant MM/GBSA and MM/PBSA rescoring. Phys. Chem. Chem. Phys. 2014, 16, 22035–22045. [CrossRef] 35. Tosco, P.; Balle, T. Open3DQSAR: A new open-source software aimed at high-throughput chemometric analysis of molecular interaction fields. J. Mol. Model. 2010, 17, 201–208. [CrossRef]

Sample Availability: Samples of the all test compounds are available from the authors.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).